Long CLIP L support for SDXL, SD3 and Flux.

Use the *CLIPLoader nodes.

Author: comfyanonymous

comfy/sd.py

@@ -445,12 +445,8 @@ class EmptyClass:
             clip_target.tokenizer = comfy.text_encoders.sa_t5.SAT5Tokenizer
         else:
             w = clip_data[0].get("text_model.embeddings.position_embedding.weight", None)
-            if w is not None and w.shape[0] == 248:
-                clip_target.clip = comfy.text_encoders.long_clipl.LongClipModel
-                clip_target.tokenizer = comfy.text_encoders.long_clipl.LongClipTokenizer
-            else:
-                clip_target.clip = sd1_clip.SD1ClipModel
-                clip_target.tokenizer = sd1_clip.SD1Tokenizer
+            clip_target.clip = sd1_clip.SD1ClipModel
+            clip_target.tokenizer = sd1_clip.SD1Tokenizer
     elif len(clip_data) == 2:
         if clip_type == CLIPType.SD3:
             clip_target.clip = comfy.text_encoders.sd3_clip.sd3_clip(clip_l=True, clip_g=True, t5=False)
@@ -475,10 +471,12 @@ class EmptyClass:
         clip_target.tokenizer = comfy.text_encoders.sd3_clip.SD3Tokenizer
 
     parameters = 0
+    tokenizer_data = {}
     for c in clip_data:
         parameters += comfy.utils.calculate_parameters(c)
+        tokenizer_data, model_options = comfy.text_encoders.long_clipl.model_options_long_clip(c, tokenizer_data, model_options)
 
-    clip = CLIP(clip_target, embedding_directory=embedding_directory, parameters=parameters, model_options=model_options)
+    clip = CLIP(clip_target, embedding_directory=embedding_directory, parameters=parameters, tokenizer_data=tokenizer_data, model_options=model_options)
     for c in clip_data:
         m, u = clip.load_sd(c)
         if len(m) > 0:

comfy/sd1_clip.py

@@ -542,6 +542,7 @@ class SD1Tokenizer:
     def __init__(self, embedding_directory=None, tokenizer_data={}, clip_name="l", tokenizer=SDTokenizer):
         self.clip_name = clip_name
         self.clip = "clip_{}".format(self.clip_name)
+        tokenizer = tokenizer_data.get("{}_tokenizer_class".format(self.clip), tokenizer)
         setattr(self, self.clip, tokenizer(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data))
 
     def tokenize_with_weights(self, text:str, return_word_ids=False):
@@ -570,6 +571,7 @@ def __init__(self, device="cpu", dtype=None, model_options={}, clip_name="l", cl
             self.clip_name = clip_name
             self.clip = "clip_{}".format(self.clip_name)
 
+        clip_model = model_options.get("{}_class".format(self.clip), clip_model)
         setattr(self, self.clip, clip_model(device=device, dtype=dtype, model_options=model_options, **kwargs))
 
         self.dtypes = set()

comfy/sdxl_clip.py

@@ -22,7 +22,8 @@ def __init__(self, tokenizer_path=None, embedding_directory=None, tokenizer_data
 
 class SDXLTokenizer:
     def __init__(self, embedding_directory=None, tokenizer_data={}):
-        self.clip_l = sd1_clip.SDTokenizer(embedding_directory=embedding_directory)
+        clip_l_tokenizer_class = tokenizer_data.get("clip_l_tokenizer_class", sd1_clip.SDTokenizer)
+        self.clip_l = clip_l_tokenizer_class(embedding_directory=embedding_directory)
         self.clip_g = SDXLClipGTokenizer(embedding_directory=embedding_directory)
 
     def tokenize_with_weights(self, text:str, return_word_ids=False):
@@ -40,7 +41,8 @@ def state_dict(self):
 class SDXLClipModel(torch.nn.Module):
     def __init__(self, device="cpu", dtype=None, model_options={}):
         super().__init__()
-        self.clip_l = sd1_clip.SDClipModel(layer="hidden", layer_idx=-2, device=device, dtype=dtype, layer_norm_hidden_state=False, model_options=model_options)
+        clip_l_class = model_options.get("clip_l_class", sd1_clip.SDClipModel)
+        self.clip_l = clip_l_class(layer="hidden", layer_idx=-2, device=device, dtype=dtype, layer_norm_hidden_state=False, model_options=model_options)
         self.clip_g = SDXLClipG(device=device, dtype=dtype, model_options=model_options)
         self.dtypes = set([dtype])
 
@@ -57,7 +59,8 @@ def encode_token_weights(self, token_weight_pairs):
         token_weight_pairs_l = token_weight_pairs["l"]
         g_out, g_pooled = self.clip_g.encode_token_weights(token_weight_pairs_g)
         l_out, l_pooled = self.clip_l.encode_token_weights(token_weight_pairs_l)
-        return torch.cat([l_out, g_out], dim=-1), g_pooled
+        cut_to = min(l_out.shape[1], g_out.shape[1])
+        return torch.cat([l_out[:,:cut_to], g_out[:,:cut_to]], dim=-1), g_pooled
 
     def load_sd(self, sd):
         if "text_model.encoder.layers.30.mlp.fc1.weight" in sd:

comfy/text_encoders/flux.py

@@ -18,7 +18,8 @@ def __init__(self, embedding_directory=None, tokenizer_data={}):
 
 class FluxTokenizer:
     def __init__(self, embedding_directory=None, tokenizer_data={}):
-        self.clip_l = sd1_clip.SDTokenizer(embedding_directory=embedding_directory)
+        clip_l_tokenizer_class = tokenizer_data.get("clip_l_tokenizer_class", sd1_clip.SDTokenizer)
+        self.clip_l = clip_l_tokenizer_class(embedding_directory=embedding_directory)
         self.t5xxl = T5XXLTokenizer(embedding_directory=embedding_directory)
 
     def tokenize_with_weights(self, text:str, return_word_ids=False):
@@ -38,7 +39,8 @@ class FluxClipModel(torch.nn.Module):
     def __init__(self, dtype_t5=None, device="cpu", dtype=None, model_options={}):
         super().__init__()
         dtype_t5 = comfy.model_management.pick_weight_dtype(dtype_t5, dtype, device)
-        self.clip_l = sd1_clip.SDClipModel(device=device, dtype=dtype, return_projected_pooled=False, model_options=model_options)
+        clip_l_class = model_options.get("clip_l_class", sd1_clip.SDClipModel)
+        self.clip_l = clip_l_class(device=device, dtype=dtype, return_projected_pooled=False, model_options=model_options)
         self.t5xxl = T5XXLModel(device=device, dtype=dtype_t5, model_options=model_options)
         self.dtypes = set([dtype, dtype_t5])
 

comfy/text_encoders/long_clipl.py

@@ -6,9 +6,9 @@ def __init__(self, embedding_directory=None, tokenizer_data={}):
         super().__init__(max_length=248, embedding_directory=embedding_directory, tokenizer_data=tokenizer_data)
 
 class LongClipModel_(sd1_clip.SDClipModel):
-    def __init__(self, device="cpu", dtype=None, model_options={}):
+    def __init__(self, *args, **kwargs):
         textmodel_json_config = os.path.join(os.path.dirname(os.path.realpath(__file__)), "long_clipl.json")
-        super().__init__(device=device, textmodel_json_config=textmodel_json_config, return_projected_pooled=False, dtype=dtype, model_options=model_options)
+        super().__init__(*args, textmodel_json_config=textmodel_json_config, **kwargs)
 
 class LongClipTokenizer(sd1_clip.SD1Tokenizer):
     def __init__(self, embedding_directory=None, tokenizer_data={}):
@@ -17,3 +17,14 @@ def __init__(self, embedding_directory=None, tokenizer_data={}):
 class LongClipModel(sd1_clip.SD1ClipModel):
     def __init__(self, device="cpu", dtype=None, model_options={}, **kwargs):
         super().__init__(device=device, dtype=dtype, model_options=model_options, clip_model=LongClipModel_, **kwargs)
+
+def model_options_long_clip(sd, tokenizer_data, model_options):
+    w = sd.get("clip_l.text_model.embeddings.position_embedding.weight", None)
+    if w is None:
+        w = sd.get("text_model.embeddings.position_embedding.weight", None)
+    if w is not None and w.shape[0] == 248:
+        tokenizer_data = tokenizer_data.copy()
+        model_options = model_options.copy()
+        tokenizer_data["clip_l_tokenizer_class"] = LongClipTokenizer_
+        model_options["clip_l_class"] = LongClipModel_
+    return tokenizer_data, model_options

comfy/text_encoders/sd3_clip.py

@@ -20,7 +20,8 @@ def __init__(self, embedding_directory=None, tokenizer_data={}):
 
 class SD3Tokenizer:
     def __init__(self, embedding_directory=None, tokenizer_data={}):
-        self.clip_l = sd1_clip.SDTokenizer(embedding_directory=embedding_directory)
+        clip_l_tokenizer_class = tokenizer_data.get("clip_l_tokenizer_class", sd1_clip.SDTokenizer)
+        self.clip_l = clip_l_tokenizer_class(embedding_directory=embedding_directory)
         self.clip_g = sdxl_clip.SDXLClipGTokenizer(embedding_directory=embedding_directory)
         self.t5xxl = T5XXLTokenizer(embedding_directory=embedding_directory)
 
@@ -42,7 +43,8 @@ def __init__(self, clip_l=True, clip_g=True, t5=True, dtype_t5=None, device="cpu
         super().__init__()
         self.dtypes = set()
         if clip_l:
-            self.clip_l = sd1_clip.SDClipModel(layer="hidden", layer_idx=-2, device=device, dtype=dtype, layer_norm_hidden_state=False, return_projected_pooled=False, model_options=model_options)
+            clip_l_class = model_options.get("clip_l_class", sd1_clip.SDClipModel)
+            self.clip_l = clip_l_class(layer="hidden", layer_idx=-2, device=device, dtype=dtype, layer_norm_hidden_state=False, return_projected_pooled=False, model_options=model_options)
             self.dtypes.add(dtype)
         else:
             self.clip_l = None
@@ -95,7 +97,8 @@ def encode_token_weights(self, token_weight_pairs):
             if self.clip_g is not None:
                 g_out, g_pooled = self.clip_g.encode_token_weights(token_weight_pairs_g)
                 if lg_out is not None:
-                    lg_out = torch.cat([lg_out, g_out], dim=-1)
+                    cut_to = min(lg_out.shape[1], g_out.shape[1])
+                    lg_out = torch.cat([lg_out[:,:cut_to], g_out[:,:cut_to]], dim=-1)
                 else:
                     lg_out = torch.nn.functional.pad(g_out, (768, 0))
             else: